Beta 1

Title Multivariate Analysis of Sensory Data
Author Le, Phucan
Supervisor Brockhoff, Per B. (Mathematical Statistics, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Master's thesis
Year 2008
Abstract In this thesis a sensory profiling data set describing the quality of fish on 17 different sensory attributes as evaluated by ten assessors is analyzed with main focus on the differences between products (fish given different feed and stored on ice a different amount of days. Especially the difference resulting from different feed is of interest. The analysis will be done by a number of multivariate methods. Comparison of the performance of the methods on the given data set is also an objective of the thesis, but only secondary. The data analytical strategy involves a descriptive statistical analysis to obtain an overview of the distribution and standard deviations of the scores for each sensory attribute along with the correlations between pairs of attributes. Especially the presence of multicollinearity is of interest as the performance of the multivariate methods depends on this. A Principal Component Analysis (PCA) is employed in order to visualize the main tendencies of variation. 3-way and 4-way univariate mixed models are analysed in order to model multivariate test statistics. 3-way and 4-way mixed model MANOVA are done together with Canonical Variate Analysis (CVA) in order to visualize the main tendencies in the same way as done with the PCA, but taking the error structure into account and with p-values for difference in products. A 50-50 MANOVA is done using the principles of both PCA and MANOVA, hereby obtaining test statistics on dimension reduced data. The statistical reliability and predictive validity of the product differences are obtained by (M)ANOVA and cross validation. Similar data structures are observed in the various multivariate (and univariate) methods with slight differences. Odor, flavor and texture attributes differentiated the fish samples and the different type of feed had an effect on the sensory evaluation of the fish.
Series IMM-M.Sc.-2008-84
Original PDF ep08_84.pdf (3.85 MB)
Admin Creation date: 2008-09-01    Update date: 2008-09-01    Source: dtu    ID: 222770    Original MXD